The present invention relates to a storage system via a network in an environment having one or more servers and storage sub systems and in particular, to a method for rearranging a logical volume.
As a storage capacity of each enterprise increases, attention is given to a technique for reducing the storage management cost. For example, by using a SAN (Storage Area Network) and a NAS (Network Attached Storage), it is possible to share a single storage device by a plurality of Application Servers, thereby localizing a storage operation job. As a result, it is possible to reduce the storage operation management cost.
In such an environment, since a single storage device is shared by a plurality of Application Servers, I/O from the servers are concentrated in the single storage device. In the SAN or NAS environment, a disc array system (storage subsystem) may be used for realizing a high-speed I/O. As the technique for realizing the high-speed I/O in the disc array system, there is RAID (Redundant Array of Inexpensive Disks). The RAID manages a plurality of discs as a single set, and operates the discs in parallel for realizing high-speed I/O. Here, this disc set will be referred to as a parity group. Moreover, in general, the parity group is divided into a plurality of volumes and each volume is allocated to a plurality of servers. Here, the volume will be referred to as a logical volume.
Thus, in the SAN and the NAS environment, when logical volumes in the parity group are allocated to a plurality of servers, a plurality of servers may share the same parity group. If I/O is concentrated in a particular parity group, the I/O of the server using the parity group is deteriorated.
Conventionally, as is disclosed in U.S. Pat. No. 6,446,161, the I/O load between the parity groups has been balanced by moving the logical volume from the parity group where I/O is concentrated to another parity group where I/O is not concentrated according to the performance information of each parity group. Here, the movement of the logical volume for balancing the I/O load between the parity groups will be referred to as rearrangement of the logical volume.
It should be noted that the I/O performance is changed by a cache. Here, however, for clarification, no mention is made for the performance change by the cache.
In general, the I/O performance required by an application using a storage sub system greatly differs according to the application processing content. For example, for a server executing a transaction processing, when the response time of each transaction is 0.5 seconds or less, the average number of accesses to the storage device per one transaction is two, and 100 transactions should be processed in one second, a high-speed I/O performance is required. On the other hand, backup processing and application such as a batch processing at night do not require a high-speed I/O performance in many cases.
Thus, the I/O performance required by the application differs according to the processing content and accordingly, in rearranging the logical volume, the I/O performance required by each application should be taken into consideration. That is, in rearrangement of the logical volume, the movement destination is not decided only from the viewpoint of balancing of the I/O load between the parity groups. When the I/O load of the parity group of the movement destination is increased by the rearrangement of the logical volume, it should also be considered whether the performance of the parity group of the movement destination can satisfy the I/O performance required by each application. Thus, the movement destination should be decided.
However, in the conventional method, the movement destination parity group is decided only according to the storage performance and no consideration is given to the I/O performance required by the application using the movement destination parity group. Therefore, in the conventional method, the parity group connected by the application requiring a high-speed performance may be selected as a movement destination. As a result, the requirement value of the I/O performance of the application cannot be satisfied and the rearrangement of the logical volume may cause a performance problem. As a result, it becomes impossible to satisfy the requirement value of the I/O performance of the application and there arises a performance problem by the rearrangement of the logical volume.